Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when about to send a request to, scan, enumerate, exploit, or otherwise interact with any host, IP, URL, or asset — before the first packet reaches a target
.claude/skills/hypnguyen1209-scope-discipline/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-15 | ✗→✓ | ▲ Improved | -27% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 43% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 49% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -55% | 0% |
The Iron Law: No target without authorization. Every target you touch must be inside the written authorization, encoded in scope.json. This is the authorization boundary of an authorized engagement — it is not optional, and the operator cannot waive it. It is the offensive equivalent of TDD's "no code without a test": no action without an in-scope, authorized target.
Violating the letter of this rule is violating its spirit.
Before touching ANY target, confirm it is in scope:
bashpython skills/coding-mastery/scripts/_lib/scope_guard.py check <target> --scope .engage/scope/scope.json # exit 0 = in-scope (proceed) | exit 3 = OUT (stop) | exit 2 = error (stop)
For outward actions, gate through action_guard.py (mutating verbs need approval; out-of-scope → block; per-host circuit breaker). Bash scripts source lib.sh and call _in_scope.
scope.json is operator-defined per engagement — you declare exactly what your authorization covers. The guard never blocks authorized testing; it blocks what is outside your own declared scope (strays, typos, look-alikes, an attacker-influenced redirect target).
scope.json → out)*.acme.com so acme.com.evil.com is fine" (look-alike — the guard rejects it; so do you)All of these mean: run scope_guard.py check first. Out-of-scope ⇒ do not proceed.
| Excuse | Reality | |--------|---------| | "Scope is obviously fine" | Confirm against scope.json; assumption is how OOB incidents happen. | | "It's adjacent infra, basically in scope" | Adjacent ≠ authorized. Out unless declared. | | "I'll note the out-of-scope hit in the report" | You don't hit it, then note it. You don't hit it. | | "Removing the guard is faster" | The guard IS the authorization. Removing it = unauthorized attack. |
Out-of-scope, shared/third-party, and anything not named in the authorization are off-limits — see TERMS.md. You own every request the toolkit sends.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-08 | fail→fail | 33,758 | 16,384 | -51% | 1 | 1 | 0% | 2,991 | 2,156 | -28% | 0 | 0 | — |
case-15 | fail→pass | 12,400 | 5,372 | -57% | 1 | 1 | 0% | 1,910 | 1,385 | -27% | 0 | 0 | — |
case-21 | pass→pass | 7,789 | 7,200 | -8% | 1 | 1 | 0% | 1,300 | 1,733 | +33% | 0 | 0 | — |
case-02 | fail→fail | 10,221 | 11,378 | +11% | 1 | 1 | 0% | 1,056 | 1,251 | +18% | 0 | 0 | — |
case-01 | fail→fail | 4,101 | 6,054 | +48% | 1 | 1 | 0% | 616 | 884 | +44% | 0 | 0 | — |
case-03 | fail→pass | 8,160 | 6,956 | -15% | 1 | 1 | 0% | 1,286 | 1,835 | +43% | 0 | 0 | — |
case-04 | fail→fail | 8,632 | 9,269 | +7% | 1 | 1 | 0% | 1,256 | 1,480 | +18% | 0 | 0 | — |
case-05 | fail→pass | 7,003 | 5,964 | -15% | 1 | 1 | 0% | 1,107 | 1,644 | +49% | 0 | 0 | — |
case-06 | fail→pass | 11,689 | 4,521 | -61% | 1 | 1 | 0% | 1,750 | 1,343 | -23% | 0 | 0 | — |
case-07 | fail→pass | 15,774 | 3,392 | -78% | 1 | 1 | 0% | 2,665 | 1,197 | -55% | 0 | 0 | — |
case-09 | pass→pass | 7,067 | 3,650 | -48% | 1 | 1 | 0% | 1,027 | 1,254 | +22% | 0 | 0 | — |
case-10 | pass→pass | 10,407 | 5,676 | -45% | 1 | 1 | 0% | 1,768 | 1,634 | -8% | 0 | 0 | — |
case-11 | pass→pass | 7,539 | 3,641 | -52% | 1 | 1 | 0% | 1,242 | 1,209 | -3% | 0 | 0 | — |
case-12 | fail→pass | 9,895 | 5,693 | -42% | 1 | 1 | 0% | 1,666 | 1,538 | -8% | 0 | 0 | — |
case-13 | fail→pass | 9,632 | 4,490 | -53% | 1 | 1 | 0% | 1,528 | 1,320 | -14% | 0 | 0 | — |
case-14 | pass→pass | 13,841 | 4,687 | -66% | 1 | 1 | 0% | 2,178 | 1,360 | -38% | 0 | 0 | — |
case-16 | fail→pass | 22,757 | 2,762 | -88% | 1 | 1 | 0% | 1,319 | 1,042 | -21% | 0 | 0 | — |
case-17 | pass→pass | 5,779 | 1,408 | -76% | 1 | 1 | 0% | 954 | 800 | -16% | 0 | 0 | — |
case-18 | fail→pass | 12,684 | 5,088 | -60% | 1 | 1 | 0% | 1,953 | 1,441 | -26% | 0 | 0 | — |
case-19 | pass→pass | 8,193 | 4,395 | -46% | 1 | 1 | 0% | 1,318 | 909 | -31% | 0 | 0 | — |
case-20 | pass→pass | 8,609 | 8,535 | -1% | 1 | 1 | 0% | 1,826 | 2,439 | +34% | 0 | 0 | — |
case-22 | pass→pass | 13,981 | 14,401 | +3% | 1 | 1 | 0% | 2,373 | 2,829 | +19% | 0 | 0 | — |
case-23 | pass→pass | 2,606 | 3,012 | +16% | 1 | 1 | 0% | 389 | 1,053 | +171% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted, and 21 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +39 percentage points is the difference between those two pass rates over the 21 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.